Skip to main content
. 2018 Jul 31;8:11489. doi: 10.1038/s41598-018-29768-x

Figure 3.

Figure 3

Tuning of regularization parameters. Normalized mean squared error (NMSE) values are plotted vs. regularization parameters. The NMSE measures the closeness of fit between the deconvolution result and a ground truth data set collected using a confocal microscope scan of the same fluorescent specimen. We evaluated how the choice of regularization parameter effects the NMSE for three deconvolution methods and three different samples with different morphologies. Because the magnitude of regularization parameters is different in different software, the parameters have been independently normalized to the range [0.0, 1.0]. We tested the following regularization parameters: 0.005, 0.010, 0.015, …, 0.100 for our method, 0, 1, 2, …, 20 for Huygens and 15, 105, 205, 305, 405, …, 1005 for Microvolution. Note that the regularization strength increases as the parameter increases in our software but decreases as the parameter increases in Huygens and Microvolution.